When Good Enough is no longer Good Enough
As a teenager I was in a band. We’d tune up and get to a point where we’d say, “That’s close enough for Rock n Roll”, which I learnt later was also an album by the Scottish band Nazareth. What the saying meant was our tuning may not be 100% accurate but it was good enough to play. The concept that things don’t have to be perfect just “good enough” has been around for many years, however as recent developments and events have shown, this is no longer the case.
The impact of online business on the high street has eroded margins to the extent that 100% accuracy is needed for forecasts. Likewise, in the Service industry, knowing exactly how long the service is needed or resources are required, has to be accurate for maximum utilisation In manufacturing, predictable production capacity can reduce finished goods buffers by more than 30%, whilst moving to a use-based preventative maintenance model, means a reduction in the frequency and duration of plant outage due to maintenance or repair. The days of the estimate on napkins or the back of an envelope is no longer relevant or acceptable in today’s business world. Good enough is no longer good enough.
For example, take a services company whose spreadsheet-based resource planning system routinely relies on 3rd party, short-term contractors or agency staff to make good the shortfall. The challenge the company faces is either build in the contingency (another “good enough” tactic) to cover these unknown costs and risk being uncompetitive, or pare it back to the minimum and risk losing money or burning out staff to cover the shortfall.
Alternatively, think of the retailer who is faced with the decision – when to discount. Discount too early and fail to optimise revenue. Wait too long and run the risk of being left with unsold stock and the prospect of having to offer even greater discounts (plus the increased cost of storage).
Here is the rub. In making our decisions and deciding what course of action to take, we set off a chain of increasingly complex events and trigger costs which we never thought to include in our original models or estimates, the consequence of which can turn a previously comfortable, good enough position, into a decidedly uncomfortable one.
The frustrating aspect is that within our organisations we have the data points we need in order to make better informed decisions. It’s just that we haven’t been able to put them together.
But this is where Machine Learning [ML] comes in. Using ML, we can start to build our models, not just using the core variables or attributes, but incorporating other, less obvious ones such as team performance e.g. which team members work well together (or more importantly which ones do not) or shortest distance travelled. As we realise the importance of the different attributes, we are able to run and re-run the model to deliver more accurate predictions about future outcomes. So, for example we can understand the importance of the level of iron found in engine oil on the performance of the engine.
Another example is a plant hire company which can run longer service intervals between it’s hired out plant, thereby increasing rental revenues at the same time as reducing maintenance costs and outage windows, ultimately moving from a fixed interval services model to a condition-based services model. The Services’ company can better schedule its resources, taking into consideration staff related issues and concerns, reducing the need for external contractors which reduces its’ costs to service the contract (thereby maximising the opportunity to increase profit) whilst delivering an improved working environment for its’ employees.
This is how we can move beyond Good Enough.
When comparing Predictive Forecasting against traditional forecasting, the predictive model will tend to be more accurate, usually by at least 10-15%. This is because the natural, optimistic tendency of humans introduces an unconscious bias into the estimates, where as the ML just relies on the data and the model (algorithm), refining its predictions as each new set of data becomes available.
For companies looking to move beyond Good Enough, ML offers a low risk, incremental approach. Starting with a simple Proof of Concept (PoC), the model can be developed and tested to create a minimal viable product (MVP). The PoC would typically take a subset of the population – a specific item of Plant or a subset of stores and prove out, comparing the outcome to previous experience. The MVP can be further developed, either by the customer or in conjunction with HSO, to ensure it can handle the full population. This incremental approach fits well with organisations looking to become agile. However, the real benefit of this approach is the initial outlay is modest (depending upon the complexity of the business) and the benefits generated can be reinvested into the next sprint.
So, whilst Good Enough is no longer good enough, HSO can show you how ML can help break through the Good Enough mentality and introduce you into the profitable, optimised operations that lie beyond.
For more information download the HSO White Paper below.